Image Processing for Adaptive Optic SystemS

Image Processing Methods for Adaptive Optic Systems

Douglas G. Currie, Petras V. Avizonis
Astro-Metrology Group, Dept. of Physics University of Maryland at College Park

The Astro-Metrology Group at the University of Maryland at College Park is currently researching the problem of generating improvements in images recorded using the already much improved Adaptive Optic telescope systems such as the Starfire Optical Range (SOR) [Fugate] and the European Space Observatory [Hubin].

The challenge:

adaptive telescopes compensate for atmospheric distortion effectively for objects that lie within the system's isoplanatic patch. For objects that lie within the field of veiw, but outside the isoplanatic patch, the Point Spread Function for the image becomes dependant upon the distance from the guide star. In this situation, it becomes diffeicult to use deconvolution methods to increase the image resolution due to this divergence of the PSF. Below are two images, the first is one which shows an image of Trapezium as seen by the SOR system at Phillips Labs, Albuquerque [Image courtesy of Bob Fugate]. The second image shows the distribution of star widths as a funciton of distance away from the guide star.

What we see in the second figure is that the widths of the stars clearly diverge from that of the guide star (at the center), which implies that deconvolution of the FOV using the guide star as the PSF will not yield good results at the edge of the image since the PSF is not valid there.

Our Approach:

currently we are working with a synthetic field which has quadratic broadening of the stars as they are seen farther from the center of the FOV.
Our approach is simple, that is to attempt to remove the broadening of the stars in the FOV, then deconvolve, then re-insert the broadening to retain the original scales of the image. A preliminary attempt at this yielded promising results, as shown in the figure below.
What we see in this result, is the + data points represent the actual widths, which we note are broadened quadratically. The * points are the measured widths after deconvolving the entire image with the central star, or the guide star. Clearly, the improvements in widths using a straight deconvolution are not as good at the edges of the FOV as they are near the center. The diamonds, on the other hand, represent the results of our Warped-Lucy Deconvolution, and it is clear that the improvements (roughly a factor of 2) in widths are uniform across the entire field, not only near the guide star. The scatter of the Warped-Lucy data points is due to shearing effects in our current procedure of removal of broadening.

The Future:

work is currently underway to reduce the shearing effects of our current procedure, then plans are to apply the procedure to the T44/SOR image at the top of this page (as well as other fields obtained by AO systems). Our main interest in performing these improvements is in order to process extended objects which have very fine spatial features such as the plasma clouds near the core of Eta Carinae, but we have determined that the best scenes to develop these tools are clusters of stars since they better show the strengths and weaknesses of various algorithms and selected parameters.

Please excuse any components of this page that do not work yet - it's all still evolving.

Return to Astro-Metrology Homepage